The composition and physical properties of the extracellular matrix (ECM) critically influence tumor progression, but the molecular mechanisms underlying ECM layering are poorly understood. Tumor–stroma interaction critically depends on cell communication mediated by exosomes, small vesicles generated within multivesicular bodies (MVBs). We show that caveolin-1 (Cav1) centrally regulates exosome biogenesis and exosomal protein cargo sorting through the control of cholesterol content at the endosomal compartment/MVBs. Quantitative proteomics profiling revealed that Cav1 is required for exosomal sorting of ECM protein cargo subsets, including Tenascin-C (TnC), and for fibroblast-derived exosomes to efficiently deposit ECM and promote tumor invasion. Cav1-driven exosomal ECM deposition not only promotes local stromal remodeling but also the generation of distant ECM-enriched stromal niches in vivo. Cav1 acts as a cholesterol rheostat in MVBs, determining sorting of ECM components into specific exosome pools and thus ECM deposition. This supports a model by which Cav1 is a central regulatory hub for tumor–stroma interactions through a novel exosome-dependent ECM deposition mechanism.
The tumor microenvironment is a key factor in multiple stages of disease progression, including local invasion and distant metastasis (Chen et al., 2015; Quail and Joyce, 2013; Quail et al., 2013). Cancer-associated fibroblasts (CAFs) are an important cell subpopulation and constitute a major source of paracrine growth factors and ECM components (Augsten et al., 2014; Kalluri, 2016). However, the molecular mechanisms driving ECM deposition are incompletely understood. Recent studies confirm that ECM deposition ultimately depends on a functional core secretion machinery (Unlu et al., 2014), but further mechanistic characterization is warranted.
Exosomes are vesicles secreted by most cell types, generated from intraluminal vesicles (ILVs) of multivesicular bodies (MVBs), membrane-bound compartments linked to the endosomal system (Colombo et al., 2014; Gruenberg and Maxfield, 1995; van Niel et al., 2006). Exosomes enable a given cell to modulate the behavior of other cell populations through the transfer of effector or regulatory molecules, often over long (interorgan) distances (Braicu et al., 2015; Raposo and Stoorvogel, 2013), and can mediate tumor–stroma communication and premetastatic niche formation (Costa-Silva et al., 2015; Hoshino et al., 2015; Peinado et al., 2012).
Exosomal cargo composition is complex, containing proteins, small ligand peptides, and regulatory nucleic acids (Colombo et al., 2014), but its regulation is poorly understood. Endosomal sorting complexes required for transport (ESCRT) machinery–dependent and ESCRT-independent mechanisms coexist (Babst, 2011; Villarroya-Beltri et al., 2014), reflecting the heterogeneity of MVBs and exosomes and their potential differential impact on health and disease. Exosome biogenesis is also modulated by the organization of membrane subdomains through lipid composition of MVB membranes (Record et al., 2018). Cholesterol is likely relevant, given its higher abundance at MVBs compared with other compartments (Möbius et al., 2003) and its involvement in the formation of high-curvature membrane structures such as caveolae and synaptic vesicles. Cholesterol, long-chain sphingolipids, and phosphatidylserine (PS) 18:0/18:1 are also significantly enriched in exosome membranes (Skotland et al., 2019; Wang et al., 2007; Wubbolts et al., 2003). ER-endosome/MVB contacts allow for the regulated transfer of proteins and lipids, especially cholesterol, and are modulated through specific protein tethers, especially members of the oxysterol-binding protein (ORP) family (ORP1L and ORP5). Cholesterol accumulation decreases ER-MVB contacts (Raiborg et al., 2015; Rocha et al., 2009). Still, we lack a complete picture of the mechanisms modulating cholesterol levels at MVBs and their impact on exosome biology.
Endomembrane trafficking is regulated by caveolin-1 (Cav1), an essential scaffolding component of invaginated plasmalemmal nanodomains (caveolae). Cav1 participates in several essential cell functions, including mechanoadaption, endocytosis, spatiotemporal control of signal transduction, cholesterol homeostasis (Parton and del Pozo, 2013; Pol et al., 2020), and interorganelle communication, being an integral component of mitochondria-associated ER membranes, a membrane contact site (MCS) of specialized ER regions and mitochondria (Sala-Vila et al., 2016). Cav1 plays multiple, yet context-dependent, roles in tumor biology and is required for ECM biomechanical remodeling by CAFs promoting tumor invasion and metastasis (Goetz et al., 2011). Exosomal secretion of Cav1 occurs (Logozzi et al., 2009), but its relevance for Cav1-dependent stromal remodeling is unknown.
Here, we demonstrate that Cav1 is a constitutive component of exosomes produced by several cell types, including primary CAFs. N-terminal ubiquitin acceptor lysines are essential for Cav1 sorting into the MVB compartment. Comparative proteomics and lipidomics revealed that Cav1 modulates exosome cargo sorting and segregation of exosome subpopulations through a cholesterol-dependent mechanism to enable incorporation of ECM components, including Tenascin-C (TnC), an important driver of tumor progression, and exosome secretion is strictly required for their deposition. Exosomes derived from Cav1 knockout (KO) fibroblasts failed to deposit ECM and promote tumor cell invasiveness. Cav1-dependent remote exosomal ECM deposition occurs in vivo, suggesting the relevance of stromal Cav1 for distant ECM niche nucleation. We propose that Cav1 is a key regulatory hub also fostering stromal environments through the modulation of MVB membrane plasticity and exosome cargo specification, which depend on cholesterol content and organization.
Cav1 redistributes to late endosome (LE)/MVB to its sorting in exosomes
Cav1 internalization is tightly regulated (del Pozo et al., 2005; Nabi and Le, 2003; Zimnicka et al., 2016). Although endocytosed Cav1 has commonly been regarded as committed to degradation (Botos et al., 2008; Hayer et al., 2010), alternative fates cannot be excluded (Mundy et al., 2012). We analyzed the trafficking dynamics of endogenous Cav1 upon treatment with sodium orthovanadate, which induces Cav1 internalization by protecting Cav1 phosphorylation at tyrosine 14 (del Pozo et al., 2005). Orthovanadate exposure increased Cav1 phosphorylation and colocalization with LEs and MVBs (anti–lyso-bis-phosphatidic acid [LBPA] and anti-CD63 immunostainings, respectively; Fig. 1 A and Fig. S1 A). Similarly, a significant fraction of Cav1 relocalized to the LE/MVB compartment upon exposure of cells to EDTA, which induces Cav1 endocytosis by promoting cell detachment (Fig. S1 B; Grande-García and del Pozo, 2008). RNAi-mediated depletion of polymerase I and transcript release factor (PTRF/Cavin-1), an essential caveolar coat component required for caveolae stabilization (Hill et al., 2008), also induced partial Cav1 relocalization—albeit below significance threshold—and increased Cav1 phosphorylation (Fig. S1 C). High-resolution microscopy confirmed the constitutive presence of Cav1 in the MVB lumen, colocalizing with the exosomal marker CD63 (Fig. 1 B). Discrete patches of Cav1 localizing to ILVs within MVBs were clearly observable by artificially enlarging endosomes (Fig. 1 C and Fig. S1 D; Trajkovic et al., 2008). These observations demonstrate that MVBs are one of the destinies of endocytosed Cav1.
Phosphorylation often triggers subsequent ubiquitination of adjacent acceptor residues (Hunter, 2007). We analyzed the subcellular distribution of Cav1 mutants in which ubiquitin acceptor lysine residues are replaced by arginine (Fig. 1 D, upper panel). Mutation of Cav1 N-terminal lysines impairs Cav1 ubiquitination (Kirchner et al., 2013). Cav1 constructs nicely decorated the outer enlarged-Rab5 endosome membrane/MVB; however, mutation of the N-terminal lysine cluster (K5-57R and K5-176R mutants) abolished Cav1 sorting into ILVs (Fig. 1 D), correlating with lower overall colocalization of these mutants with CD63 (Fig. S1 E). In contrast, mutation of C-terminal lysine residues (K65-176R) did not affect the ability of Cav1 to localize in MVBs (Fig. 1 D). Phosphorylation was not significantly altered in Cav1 mutants (Fig. 1 E), reinforcing the notion of a two-step mechanism for Cav1 sorting into MVBs: phosphorylation first elicits Cav1 internalization toward the outer MVB membrane, and ubiquitination acts as a checkpoint for Cav1 sorting into ILVs.
To assess the exosomal sorting of the ILV-localized Cav1 pool, we analyzed density gradient–purified exosomes from different cell lines. Endogenous Cav1 mainly partitioned with Tsg101, confirming that exosomes can carry Cav1 (Fig. 1 F). Cav1 sorting to exosomes was enhanced in PTRF-depleted mouse embryonic fibroblasts (MEFs; Fig. S1 F). In contrast, N-terminal ubiquitination mutants exhibited impaired exosomal sorting (Fig. 1 G). These data suggest that a significant proportion of internalized Cav1 is sorted to the MVB lumen in a ubiquitin-dependent manner, favoring its incorporation into exosomes. This process is constitutive in all Cav1-expressing cell lines tested.
Cav1 regulates exosomal biogenesis through modulation of MVB cholesterol content
The regulated composition and organization of MVB membranes yield different ILV and exosome populations (Colombo et al., 2014). We explored whether Cav1, a regulator of endomembrane compartments (Pol et al., 2020), plays an active role in exosome biogenesis. Cav1KO fibroblasts were not impaired for exosome biogenesis and displayed higher rates of exosome secretion, as determined by Western blot of total purified exosomes from equal cell numbers (Fig. 2, A and B, upper panel in A), sucrose density gradient sedimentation (Fig. 2 A, lower panel; and Fig. 3 D), and physical characterization (Nanosight technology, transmission EM [TEM]; Fig. 2, C and D; Dragovic et al., 2011). We identified two exosome subpopulations of different sizes in WT fibroblasts, whereas Cav1KO fibroblasts produced a homogeneous population of small exosomes (Fig. 2, C and D).
Cav1 binds cholesterol and organizes specialized membrane organization domains (Murata et al., 1995; Pol et al., 2020). We studied whether Cav1 promotes exosome heterogeneity by modulating cholesterol levels at MVBs. While the number of LBPA-positive structures did not differ between WT and Cav1KO fibroblasts, MVB area and filipin intensity (a proxy for mean cholesterol content) were significantly higher in Cav1KO cells (Fig. 2 E). This phenotype was recapitulated by siRNA depletion of Cav1 in WT cells (Fig. S2 A). Treatment with U18666A, which promotes cholesterol accumulation in the LE/MVB compartment (Cenedella, 2009), also increased exosome secretion while reducing exosome heterogeneity in WT cells (Fig. 2 F).
A major share of MVB cholesterol is contained in ILVs (Möbius et al., 2003). We confirmed this cholesterol accumulation in enlarged endosomes by overexpressing constitutively active Rab5(Q79L) (Fig. S2 B). To confirm whether altered cholesterol content in Cav1KO MVBs induces exosome membrane reorganization, we isolated cold-detergent–resistant membranes (exosome DRMs) from exosomes produced by either WT cells (control or U18666A treated) or Cav1KO cells. Flotillin-1, a marker for exosomes and liquid ordered domains, showed a wider distribution in Cav1KO cells (fractions 1 to 8) or U18666A-treated WT cells (fractions 1 to 6); in contrast, flotillin accumulation in untreated WT cell–derived exosomes was restricted to fractions 1 to 3 (Fig. 2 G), reflecting the increased highly ordered membrane content of Cav1KO-derived exosomes. Lipidomic analyses of exosome preparations confirmed a significant enrichment in PS—in particular PS (18:0/18:1) species—in agreement with previous data (Skotland et al., 2019). Of note, exosomes derived from WT cells invariably exhibited higher relative amounts of phosphatidic acid, phosphatidylinositol, and LBPA (which is known to play a relevant role in ILV biogenesis and dynamics; Cenedella, 2009; Kowal et al., 2014; Subra et al., 2007) compared with Cav1KO exosomes, which conversely displayed an increased amount of cholesterol (Fig. S2 C), in agreement with the observed accumulation of cholesterol at LE/MVB compartments of Cav1KO cells. These results suggest that Cav1 regulates exosome biogenesis by acting as a “cholesterol rheostat” at LE/MVBs, conferring plasticity to this compartment and enabling the segregation of different exosome populations.
Cav1 promotes the sorting of specific ECM cargoes into exosomes
We quantitated the proteomes of WT and Cav1KO fibroblasts and their corresponding exosome preparations. This analysis identified a subproteome that is equally abundant in WT and Cav1KO whole-cell lysates, but whose components are differentially sorted into exosomes depending on the Cav1 genotype (Fig. 3 A, left). With a threshold of |Zq| > 1.5, in Cav1KO exosomes 152 proteins were more abundant and 163 less abundant than in WT exosomes. Interaction network and functional annotation enrichment analysis revealed that WT-derived exosomes were significantly enriched in ECM components (including TnC, fibronectin [FN], nidogen, emilin, EDIL3, and heparan sulfate proteoglycans) and endosomal/endomembrane system components (Fig. 3 A, right upper panel; Fig. 3 B; Fig. S2 D; and Data S1). Conversely, a marked enrichment for DNA/RNA binding and chaperone proteins such as histones in Cav1KO exosomes was observed (Fig. 3 A, right lower panel; Fig. 3 B; Fig. S2 D; and Data S1). Consistent with our data supporting increased cholesterol content in Cav1KO MVBs, various annexin family members were substantially enriched in Cav1KO exosomes, despite a decreased abundance of these proteins in Cav1KO whole-cell lysates (Fig. 3 C). Annexins are Ca2+-dependent proteins that bind cholesterol-rich membranes, providing lipid-clustering activity, and may constitute a mechanism to accommodate cholesterol excess in Cav1KO exosomes (Lizarbe et al., 2013).
Differential exosome sorting of selected proteins was validated by Western blot analysis of exosome fractions (Fig. S2 G and Fig. S3 I) and iodixanol density gradients (Fig. 3 D). As shown in Fig. 3 D, the distributions of exosome markers and cargoes markedly correlated. A particular difference was observed for TnC (Zq = 3.02), an ECM protein involved in development, stem cell niche formation, and aggressive tumor metastasis (Lowy and Oskarsson, 2015; Midwood et al., 2016; Midwood and Orend, 2009). TnC was prominent in WT-derived exosomes but virtually absent in Cav1KO exosomes (Data S1). TnC colocalized with Cav1 within MVBs in fibroblasts and U251 cells (a TnC-secreting glioblastoma cell line; Fig. S2 E; Brösicke et al., 2013). Knockdown (KD) of Cav1 across different tumor cell lines (breast, glioblastoma, and melanoma) recapitulated the phenotype of blunted TnC secretion observed in Cav1KO MEFs (Fig. S2, H and I).
This supports that Cav1 modulates exosome biogenesis and ECM cargo sorting across different cell types and is a prominent regulator of exosome-mediated intercellular communication.
Exosome biogenesis/secretion modulates ECM deposition
We examined whether exosome-delivered ECM components can initiate de novo ECM deposition. Cav1-deficient fibroblasts deposited less TnC fiber matrix than WT cells (Fig. 4 A). TnC fiber deposition was also reduced in U18666A-treated cells (Fig. 4, B and C; and Fig. S3 B). Reduced TnC deposition correlated with intracellular TnC accumulation in Cav1-deficient cells; this effect was less evident in U18666A-treated cells, probably reflecting a higher TnC degradation rate (Fig. 4 B). TnC was less abundant in exosomes derived from Cav1KO and Cav1KD MEFs or U18666A-treated WT MEFs (Fig. 4 D and Fig. S2 F). We used two different exosome secretion inhibitors to assess the contribution of exosomes to extracellular TnC fiber deposition: dimethylamiloride (dmA; a H+/Na+ and Na+/Ca2+ channel inhibitor) and GW4869 (a neutral sphingomyelinase [nSMase]-2 inhibitor; Savina et al., 2003; Trajkovic et al., 2008). Net exosome release by WT fibroblasts was reduced by GW4869 and to a lesser extent by dmA (Fig. S3 A), correlating with impaired TnC fiber deposition (Fig. 4 E). Both treatments led to intracellular accumulation of TnC aggregates (Fig. 4 E and Fig. S3 C). Deposition of FN, another key ECM component underrepresented in Cav1KO exosomes (Fig. 3 B, Fig. S3 I, and Data S1), was also reduced by exposure to exosome secretion inhibitors, with GW4869 having a much stronger effect (Fig. 4 F), although no significant differences could be observed upon U18666A treatment (Fig. S3 K). Relevance of exosome secretion for ECM deposition was also confirmed in primary CAFs, in which exosome release was largely inhibited by GW4869, whereas dmA had virtually no effect (Fig. S4 A). Accordingly, TnC and FN deposition was specifically affected by nSMase2 blockade, whereas dmA had no effect (Fig. S4 B and Fig. 4 C). Neither GW4869 nor dmA affected ER-Golgi homeostasis in MEFs or CAFs (Fig. S3 D, Fig. 3 E, Fig. S4 D, and Fig. 4 E), ruling out compromised ER-Golgi secretory function as a potential cause of defective deposition of ECM components.
ESCRT machinery-regulated trafficking and ceramide biogenesis pathways generate exosomes with different cargoes (Colombo et al., 2014). We analyzed the effect of transient KD of either Tsg101 (a core regulator of the ESCRT-dependent pathway) or nSMase1 and 2 (key regulators of ceramide metabolism) on TnC matrix deposition (Fig. S3, F and G). Consistent with small inhibitor experiments (Fig. 4 E), KD of either pathway impaired TnC deposition, with ceramide synthesis disruption yielding stronger effects (Fig. 4 G and Fig. S3 H). These treatments also reduced FN deposition (Fig. S3 J). Thus, exosomes are a general ECM deposition mechanism, and ceramide-dependent exosome biogenesis is a major limiting step in this process.
TnC incorporation into exosomes has an intracellular origin and depends on cholesterol modulation
To rule out an external origin of exosome-incorporated TnC through endocytic trafficking, we developed an assay based on cell-derived matrices (CDMs). CDMs were generated by decellularization of ECMs synthesized in vitro by WT or TnCKO MEFs. We confirmed that TnCKO cells were specifically impaired for TnC deposition; other components, such as FN, did not exhibit such dramatic dependence and showed a slight up-regulation, likely because of the existence of alternative compensatory routes (Fig. 5 A and Fig. S5 A; see also Fig. S3 I). TnCKO or WT cells were then seeded on each of these CDMs (Fig. 5 B), and TnC internalization was determined. No intracellular TnC was detected inside TnCKO fibroblasts grown on TnC-rich CDMs, thus excluding endocytosis as an active TnC pool for exosome cargo sorting.
To exclude the possibility that TnC detected in purified exosome fractions came from passive, nonspecific binding of extant extracellular TnC fibers to secreted exosomes, purified exosomes from TnCKO cells were incubated with TnC-rich matrix produced by WT cells (Fig. 5 C and Materials and methods section). We failed to detect TnC protein in TnCKO exosomes after their incubation with TnC-rich matrix (Fig. 5 C).
We then separated MVBs from the ER and Golgi compartments (which fractionate together). In untreated WT cells, TnC cosedimented both with MVB and ER markers (Fig. 5 D); however, upon treatment with GW4869, TnC was largely excluded from MVB fractions, a distribution shift recapitulated in Cav1KO cells (Fig. 5 D). Accordingly, a major fraction of cellular TnC colocalized with the ER in Cav1KO cells or GW4869-treated WT cells (Fig. 5 E and Fig. S4 F). This colocalization was enhanced upon chloroquine supplementation (Fig. S4 F), suggesting that defective transfer of TnC from the ER to MVBs results in ER retention of TnC and subsequent targeting for lysosomal degradation, supporting that lysosomal degradation enables TnC removal when its exosome sorting fails. Degradation rate differences could explain the variability in TnC accumulation in Cav1KO cells and WT cells treated with exosome inhibitor drugs and the lack of ER stress upon severe disruption of exosome biogenesis and secretion.
Thus, the presence of TnC in exosomes reflects true regulated intracellular sorting of TnC into MVBs, likely through a mechanism involving ER-MVB contact sites. This exosomal TnC pool is sourced in the cell from de novo translation and not from endocytosed material.
MCSs are small, specialized domains through which different membrane compartments communicate, allowing the active transfer of ions and lipids, including cholesterol in the case of ER-endosomal MCSs (summarized in a graphical scheme in Fig. S3 L; Rocha et al., 2009). We hypothesized that ER-MVB MCS disruption alleviates cholesterol accumulation in Cav1KO cells and restores the impaired sorting of ECM proteins onto exosomes. We knocked down ORP1L and vesicle-associated protein A (VAPA), as these components of the ER-endosome tether complex VAPA:ORP1L are relevant regulators of ER-to-endosome cholesterol transfer. Disruption of these contact sites in Cav1KO cells significantly reduced LE/MVB size and filipin staining intensity (Fig. 6 A and Fig. S5 D). Cholesterol content reduction upon disruption of these contact sites correlated with a rescue of TnC incorporation into MVBs (Fig. 6 B) and extracellular TnC fiber deposition to levels comparable to those observed in WT cells (Fig. 6 C).
To further characterize the role of endosomal cholesterol accumulation in the sorting of ECM proteins onto exosomes, we assessed the impact of loading WT cells with exogenous low-density lipoprotein (LDL)-cholesterol on TnC trafficking into LE/MVBs. Importantly, cholesterol-loaded cells showed a drastic reduction in the accumulation at LE/MVBs and extracellular deposition of TnC (Fig. S4, G and I).
Thus, the control of cholesterol content in the endosomal compartment is a key mechanism regulating exosome biogenesis and exosomal cargo specification, of which Cav1 is a relevant player.
Cav1-dependent exosome biogenesis directly drives ECM deposition and subsequent increase in tumor cell invasiveness.
We compared the ability of exosomes purified from either WT or Cav1KO MEFs to deposit TnC on 2D cultures of the noninvasive breast cancer tumor cell line MDA-MB468, which naturally expresses negligible levels of Cav1 and TnC (Kang et al., 2016). After 24 h, tumor cells internalized prelabeled WT and Cav1KO exosomes with similar efficiency, as shown by similar accumulation of the general membrane dye PKH67 (Fig. 7 A, green). However, only WT exosomes led to detectable deposition of TnC in areas close to the periphery (Fig. 7 A, in red). In contrast, MDA-MB468 cultures exposed to Cav1KO exosomes showed TnC deposition levels comparable to those displayed by control cultures (exposed to vehicle). Exosomes thus have an intrinsic ability to deposit specific ECM material in vitro that requires Cav1-dependent regulation of exosome biogenesis.
Cav1-dependent exosomal deposition of TnC induced important morphological changes in MDA-MB468 tumor cells, eliciting cell-cell adhesion loss and internalization of the epithelial marker E-cadherin, associated with increased protrusiveness and emission of filopodia (Fig. 7 B). A similar phenotype was observed when using CAF lines transduced with lentiviral shRNA-encoding vectors (Fig. S5 C). Accordingly, MDA-MB468 cells exhibited increased migration, as assessed by wound healing and transwell migration assays (Fig. S5, F and G). Interestingly, exosomes purified from MDA-MB231 tumor cells, which contain low TnC compared with exosomes derived from WT fibroblasts despite having comparable amounts of Cav1 protein, failed to up-regulate protrusiveness and motility of MDA-MB468 cells (Fig. S5 B; Fig. S2, G and H; and Data S2). Thus, exosomal TnC/ECM cargo derived from stromal fibroblasts, and not exosomal Cav1 per se, drives the induction of invasive phenotypes in tumor cells.
We studied the ability of exosomes to deposit ECM on MDA-MB468 spheroids, scaffold-free 3D assemblies in which cells self-organize ECM and engage in intercellular communication and which have been validated as superior in vitro experimental proxies of live tissues for many applications (Fennema et al., 2013; Zanoni et al., 2016). WT-derived TnC-containing exosomes, but not exosomes derived from Cav1KO or Cav1KD fibroblasts (i.e., TnC depleted), efficiently deposited fibrillar TnC structures within cell spheroids (Fig. 7 C). 3D reconstruction demonstrated that TnC fibrils were preferentially located in the interstitial space (Fig. S5 E).
To characterize Cav1-dependent, exosome-induced phenotypic changes in tumor cells in this setting, spheroids generated in the presence of exosomes were plated and cells were allowed to migrate for 72 h. WT exosomes promoted significantly more migration than exosomes derived from Cav1KO MEFs (Fig. 7 D). To confirm the role of TnC in this process, we tested the effect of exosomes derived from TnCKO MEFs (Fig. 7 G and Fig. S5, H and I). TnCKO exosomes were poor stimulators of cell migration, supporting that the observed phenotypic impact derives largely from the sorting of specific ECM cargoes to exosomes.
We next examined whether exosomes increased tumor cell invasiveness. Cav1KO exosomes consistently exhibited a blunted ability to induce invasiveness in Matrigel invasion assays (Fig. 7 E) or 3D collagen matrix invasion assays (Fig. 7 F), similar to the reduced invasion promoted by exosomes derived from TnCKO cells (Fig. 7 G). These results further support the relevance of Cav1- and TnC-dependent exosome-induced ECM deposition and phenotypic alteration for tumor progression.
Cav1WT MEF-derived exosomes create TnC-rich niches in vivo
To evaluate whether exosomes deposit TnC in vivo over large (interorgan) distances, exosomes purified from WT or Cav1KO fibroblasts were fluorescently labeled and injected into the caudal vein of TnCKO mice every day for 1 wk (Fig. 8 A). Histopathological analysis revealed specific exosome accumulation in liver and, to a lesser extent, in lung (Fig. 8 B and Fig. S5 J). However, associated TnC deposition was only consistently detectable in liver and lung of mice injected with WT exosomes, contrasting with the residual TnC staining in the organs of animals injected with Cav1KO exosomes or vehicle (PBS), despite Cav1KO exosomes being able to reach this organ with comparable efficiency. Thus, Cav1 has a key role in exosome-dependent ECM deposition in vivo. Because TnC is a positive regulator of metastasis through the assembly of favorable distant microenvironments, Cav1 might specifically modulate cancer outcome through exosome-dependent stroma–tumor communication.
Exosomes constitute a means by which tumor cells and associated stroma not only communicate locally but also influence and transform distant locations or “niches” and even modulate organismal responses systemically (Costa-Silva et al., 2015; Hoshino et al., 2015; Peinado et al., 2012). Despite their ubiquity, the mechanisms underlying exosome biogenesis, cargo sorting, and function remain elusive. Here, we identify Cav1 as a modulator of exosome biogenesis through the regulation of cholesterol content in MVBs, which allows the generation of different exosome subpopulations and exosome cargo specification (mainly specific ECM proteins) with subsequent functional impact. Our data demonstrate that the presence of Cav1 within exosomes is favored by its endocytosis through Tyr14 phosphorylation. This promotes Cav1 targeting to the MVB outer membrane, where a second mechanism is activated that requires the presence of ubiquitination-targeting N-terminal domain lysine residues; this second mechanism sorts Cav1 into ILVs for exosome generation. Cav1 would act as a “cholesterol rheostat” in MVBs, thus increasing the plasticity of this compartment and enabling the segregation of different exosome populations. Cholesterol titration in this compartment could also determine exosome cargo specification by modulating ER-MVB contacts. Cav1-dependent exosomal ECM deposition elicits both local ECM deposition and the generation of distant ECM-enriched regions, supporting a novel role of Cav1 in stromal remodeling. This exosomal ECM deposition promotes protrusiveness and active migration of targeted tumor cells by increasing cell scattering, suggesting that these mechanisms contribute to tumor cell invasiveness and favor the nucleation of new ECM stromal niches (Fig. 8 C).
TnC is a key component involved in promoting exosome-dependent tumor invasion. TnC matrix deposition by Cav1-expressing fibroblasts strictly required intact exosome biogenesis and secretion. Our data support a model whereby the ER/Golgi compartment is a major direct supplier of TnC to exosomes and rule out incorporation from routes bypassing de novo synthesis, such as internalization from the extracellular space. Supported by studies performed in breast cancer cell lines (Campos et al., 2018), our results demonstrate that Cav1 regulates a general mechanism for exosome cargo specification, especially relevant for TnC sorting and secretion. Our results suggest that Cav1-dependent exosomal TnC secretion has a dual effect on tumor progression (Fig. 7): (1) local deposition of TnC at the tumor-invasive front presumably induces tumor cell migration away from the tumor mass (see Fig. 7), as observed upon direct addition of recombinant soluble TnC (Nagaharu et al., 2011); additionally, (2) Cav1-dependent exosome–mediated TnC deposition over large distances might contribute to the “priming” of stromal niches before the arrival of cancer cells, forming a “landing dock” for subsequent metastatic growth (Costa-Silva et al., 2015; Sceneay et al., 2013). We show that IV injection of exosomes derived from WT fibroblasts, but not from Cav1KO cells, can efficiently promote direct TnC–matrix deposition in distant organs (Fig. 8). Further investigation is required to confirm whether these ECM stromal niches have an impact on metastasis in vivo.
A pathway for caveolae internalization (Chaudhary et al., 2014) has been described in which caveolae are disassembled and Cav1 is internalized and ubiquitinated, followed by accumulation at the internal membrane of LEs/MVBs (Hayer et al., 2010; Hill et al., 2008). Interestingly, Cav1 is also targeted to LEs by perturbations of lysosomal pH and changes in cholesterol content, with no changes in Cav1 degradation rate, raising the possibility of alternative fates for this Cav1 pool (Mundy et al., 2012; Pol et al., 2020). The changes in Cav1-MVB colocalization we observed (Fig. 1 and Fig. S1) are compatible with a scenario whereby a significant proportion of internalized Cav1 is sorted into MVBs, in part to be incorporated into exosomes. However, at present we cannot rule out that other intracellular compartments, such as the ER, may also be a source of exosomal Cav1.
Cav1 binds cholesterol with high affinity (Murata et al., 1995), and its ability to move between different compartments might contribute to the regulation of cholesterol fluxes in cells (Pfrieger and Vitale, 2018; Pol et al., 2005). Our data support that Cav1 dictates the levels of free cholesterol in LE/MVB membranes, conferring the required plasticity for the generation of ILVs of varying size and favoring the sorting of ECM components onto exosomes for their secretion (Fig. 2, Fig. S2, and Fig. S3). Furthermore, artificial manipulation of cholesterol levels at LE/MVB, either by disrupting ER-endosome MCSs (Fig. 6 and Fig. S5 D), by pharmacologically blocking cholesterol trafficking at the endosomal compartment (Fig. 2; Fig. 4, B and C; Fig. S2 F; and Fig. S4 H), or by exogenous loading with LDL-cholesterol (Fig. S4, G and I), had a profound effect on the intracellular trafficking and exosomal secretion of ECM exosome cargo. This suggests a model whereby the control of cholesterol levels at endosomes/MVBs is integral to the regulation of exosome-mediated ECM deposition, as recently described for other signaling events and functions at these intracellular compartments (Lim et al., 2019; Eden et al., 2016). Of note, Cav1 at ER-mitochondria MCS regulates the recruitment of sterol metabolism complexes and cholesterol accumulation (Sala-Vila et al., 2016). Further investigation will be required to better define the mechanistic details of the interplay between cholesterol trafficking and exosome biogenesis and downstream functions, of which Cav1 appears to be a key player.
Interestingly, among identified Cav1-dependent exosomal cargoes, collagens are less abundant than other ECM components. Collagen secretion requires specific assembly mechanisms operating through the COPII biosynthetic/secretory machinery (Malhotra and Erlmann, 2015) and is thus unavailable to the exosome pathway (Fig. 3, Fig. 4, and Fig. S3). These pathways might in fact be oppositely regulated by Cav1, since a negative correlation has been described between Cav1 expression and collagen deposition in chronologically aged skin and during fibrosis (Gvaramia et al., 2013; Lee et al., 2015). Supporting our observation that exosomes can act as carriers of some ECM proteins, a recent study demonstrated exosomal secretion of FN through its binding to integrin α5β1; however, the function and underlying mechanism of this sorting were not determined (Lobert et al., 2010; Sung et al., 2015). Our data point to incorporation of TnC into ILVs through its sorting from the biosynthetic route to the MVB compartment, excluding the possibility of an extracellular origin through internalization or by direct binding to exosomes once they are released (Fig. 5).
Previous studies demonstrated high levels of Cav1-expressing exosomes in the plasma of melanoma patients (Logozzi et al., 2009). We propose that relative levels of Cav1-expressing exosomes could be a valuable prognostic marker of metastasis and tumor malignancy, as well as other cholesterol-related diseases. Our findings suggest new potential opportunities for diagnosis, prevention, and therapeutic intervention through exosomal tumor–stroma communication.
Materials and methods
Antibodies and reagents
Monoclonal anti-Cav1 (#3267) and anti-Alix (#2171) antibodies were from Cell Signaling Technology. Anti-LBPA was from Echelon (Z-SLBPA). Anti-CD63 (ab1318), antiflotillin (ab13349), antiperiostin (ab14041), anticalreticulin (ab2907), anti-PTRF (ab48824), and anti-Tsg101 (ab83) were from Abcam. Anti-GM130 (610823), anti–py14-Cav1 (611339), anticadherin (610181), and antivimentin (550513) were from BD Transduction. Antitubulin (T-9026), anti-FN (F3648), anti-Tenascin (Clone Mtn-12, T3413), and anti–smooth muscle actin (A5228) were from Sigma. Anti-Tenascin was also from Millipore (AB19011). Anti-nSMase1 (sc-377135) and anti-nSMase2 (sc-166637) were from Santa Cruz. Alexa Fluor 488-, 546-, and 647-conjugated phalloidin were from Life Technologies. GW4869 (567715) was from Calbiochem, and dmA (A4562) and U18666A (U3633) were from Sigma. Matrigel (354230) was from Becton Dickinson, and Collagen I from rat tail (354249) was from Corning. Filipin was from Sigma (F4767).
MEFs derived from Cav1WT and Cav1KO littermate mice used throughout this study have been previously described (del Pozo et al., 2005). TnCWT and TnCKO MEFs were provided by our collaborator Dr. Gertraud Orend. PTRFKO cells were provided by Dr. Parton (Institute for Molecular Bioscience, University of Queensland, St. Lucia, Australia) and were reconstituted with the retroviral expression vector pMIGR1. COS7 cells were from ATCC. MEFs and COS7 cells were cultured in DMEM supplemented with 10% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin at 37°C in a humidified 5% CO2–95% air incubator. The breast cancer–associated fibroblast line CCD-1069S and the breast tumor cell line MDA-MB468 were from ATCC and were cultured in DMEM/F-12 supplemented with 10% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin. CCD-1069S fibroblasts stably expressing shRNA targeting Cav1 or a scramble siRNA were generated by lentiviral infection as reported in Goetz et al. (2011). All the cell lines were routinely checked for mycoplasma infection.
siRNA silencing was performed using 20 nM siRNA and Lipofectamine RNAiMAX Reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol, and cells were cultured for 2–3 d. siRNAs targeting mouse Tsg101, nSMase1, and nSMase2 were SMARTpool ON-TARGET plus reagents from Dharmacon (Tsg101 Cat. L-049922-01-0005; nSMase1 or Smpd2 Cat. L-044206-01-0005; and nSMase2 or Smpd3 Cat. L-059400-01-0005). The control siRNA was ON-TARGET plus nontargeting pool (Dharmacon Cat. D-001810-10-05). Cav1 was silenced with an siRNA from Dharmacon (seq. 5′-GAGCUUCCUGAUUGAGAUU-3′) and an enzymatically prepared siRNA from Sigma (EHU031251). esiRNAs targeting the VAPA:ORP1L tether complex were purchased from Sigma (EMU053971 and EMU012501, respectively).
Isolation and characterization of exosomes released by fibroblasts
Exosomes were isolated from cultured fibroblasts grown in exosome-free culture medium. To remove detached cells, conditioned medium was collected and centrifuged at 300 g for 10 min at 4°C. The supernatant was collected and centrifuged at 2,000 g for 20 min at 4°C. The supernatant was then centrifuged at 10,000 g at 4°C for 30 min to completely remove contaminating apoptotic bodies, microvesicles, and cell debris. The clarified medium was then ultracentrifuged at 110,000 g at 4°C for 70 min to pellet exosomes. The supernatant was carefully removed, and crude exosome-containing pellets were washed in ice-cold PBS. After a second round of ultracentrifugation, the resulting exosome pellets were resuspended in the desired volume of PBS.
For further purification, exosomes were ultracentrifuged on a discontinuous sucrose gradient including sucrose concentrations of 0.25 M, 0.5 M, 0.8 M, 1.16 M, 1.3 M, and 2 M in 20 mM Hepes, pH 7.4. Exosome samples were laid on the bottom of the gradient in the 2-M sucrose fraction, followed by centrifugation at 200,000 g for 18 h. 11 individual 1-ml gradient fractions were manually collected. Fractions were diluted in PBS and centrifuged at 110,000 g for 1 h at 4°C, and the resulting pellets were resuspended and analyzed. Alternatively, an equal volume of cold acetone was added to each fraction, and the proteins were precipitated for 2 h at −20°C. Protein pellets were collected by centrifugation at 16,000 g for 10 min and air-dried to eliminate acetone traces. The protein precipitates were monitored by Western blot for the expression of the exosomal markers Alix and Tsg101.
Alternatively, a discontinuous iodixanol gradient was used. To prepare the discontinuous iodixanol gradient, 40% (wt/vol), 20% (wt/vol), 10% (wt/vol), and 5% (wt/vol) solutions of iodixanol were made by diluting stock solutions of Optiprep (60% [wt/vol] aqueous iodixanol) with 0.25 M sucrose/10 mM Tris, pH 7.5. The gradient was formed by adding 3 ml of 40% iodixanol solution to a 14 × 89–mm polyallomer tube (Beckman Coulter), followed by careful layering of 3 ml each of 20% and 10% solutions and 2.5 ml of 5% solution. Exosome preparation was overlaid onto the top of the gradient and centrifugation performed at 100,000 g for 18 h at 4°C. 12 individual 1-ml gradient fractions were collected manually from the top to the bottom. Fractions were diluted with 1 ml PBS and precipitated with 2 ml cold acetone for 3 h at −20°C. Protein pellets were collected by centrifugation at 16,000 g for 10 min and air-dried to eliminate acetone traces. The protein precipitates were monitored by Western blot for the expression of exosomal markers Alix, Tsg101, flotillin-1, and the negative control GM130 (Golgi).
In all experiments, the exosomes used corresponded to the total exosome pellets resulting from the serial centrifugation steps; usually, 106 exosome particles were added per cell. Exosome uptake was performed in exosome-free medium.
Exosome concentrations and size distributions were determined by Nanoparticle Tracking Analysis (Nanosight).
Western blot analysis
Western blot samples were prepared with conventional 5× Laemmli sample buffer brought to 1× when mixed with the sample. The 5× Laemmli sample buffer is composed of 60 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5% β-mercaptoethanol, and 0.01% bromophenol blue.
Cell or exosome proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes. Blots were incubated for 2 h in PBS containing 5% skimmed milk and overnight with primary antibodies (typically diluted 1:1,000). After incubation (2 h) with horseradish peroxidase–conjugated goat anti-rabbit or goat anti-mouse antibody, signal on washed blots was detected by enhanced chemiluminescence (GE Healthcare). Band intensities were quantified with Image J software.
Intracellular lysate analysis
To prepare lysates containing only proteins from the cell interior, plated cells were treated with trypsin to eliminate proteins in the extracellular space. Detached cells were washed with PBS and resuspended in loading buffer for Western blot analysis. For total lysates comprising cell and ECM proteins, cells were scraped directly from the plate in loading buffer.
PKH67 labeling of exosomes
Purified exosomes were labeled with the green fluorescent dye PKH67 (Sigma; MIDI67) according to the manufacturer’s protocol. After staining, unbound PKH67 was removed with columns from Thermo Fisher Scientific (4484449).
Exosomes were visualized by TEM according to the method of Théry et al. (2006). The exosome suspension was fixed in 2% PFA and transferred to formvar/carbon–coated EM grids. After 20 min, grids were placed sample-face down for 2 min in a 100-μl drop of PBS on a sheet of parafilm. Grids were then transferred to 1% glutaraldehyde for 5 min and then washed with distilled water. Samples were contrasted in 2% uranyl acetate and examined by TEM.
Spheroid formation assay
After trypsinization, MDA-MB468 cells were resuspended in exosome-free DMEM, counted, and adjusted to the desired concentration (usually 5 × 103 cells/ml). A 20-µl final volume was prepared with 5× methyl cellulose prepared in exosome-free medium, the MDA-MB468 cell suspension, and the desired number of exosomes (usually 1 × 106 particles). The mixture was carefully deposited in a drop on the inner surface of a 100-mm dish lid. The plate lid was placed on a plate containing 10 ml PBS to humidify the culture chamber. Under gravity, cells aggregated at the bottom of the hanging drop. After 24 h, the resulting cell aggregates were lifted with a pipette. For 2D spheroid migration assays, the aggregates were seeded on a 96-well plate to monitor cell migration from the spheroid by phase-contrast time-lapse video microscopy. Alternatively, aggregates were seeded on collagen or Matrigel gels as described below. In all assays, no additional exosomes were added to the medium apart from the exosomes used during spheroid generation.
High-density rat-tail type I collagen (Corning) was diluted to 2 mg/ml. This solution self-polymerizes to form a gel after 20–30 min at 37°C. Tumor cell spheroids were transferred to 25 µl unpolymerized collagen in the wells of a μ-slide angiogenesis plate (ibidi). After 30 min, 40 µl medium was added over the top of the gel. Medium was replaced every 48 h. Spheroids were imaged every 24 h for the duration of the experiment. Spheroid diameter was measured from captured images with ImageJ.
For Matrigel embedding assays, spheroids were transferred to a 25-µl 1:1 mix of Matrigel and medium.
Wound closure assay
For the wound healing assay, confluent monolayers were scraped with a 0.1–2-µl pipette tip in the presence of mitomycin C, and wound closure was monitored by capturing images at 10-min intervals.
3D matrices reminiscent of in vivo ECM were prepared as published (Beacham et al., 2007; Castelló-Cros and Cukierman, 2009; Goetz et al., 2011). Briefly, 2.5 × 105 cells/ml were plated on chemically cross-linked gelatin on tissue culture dishes or coverslips and maintained in confluence for 6 d. Cells were supplemented every 48 h with 50 mg/ml fresh L-ascorbic acid to stabilize ECM components, thus facilitating collagen production and polymerization. The resulting 3D cultures were checked for quality by immunofluorescence, or cells were removed from the matrix by alkaline detergent extraction, yielding cell-free 3D matrices for further analysis.
For analysis of exosome binding to preexisting CDMs, CDMs from TnCWT and TnCKO MEFs were obtained as previously indicated. Cell-free 3D matrices were scrapped in 1 ml PBS and homogenized passing several times through a syringe (0.5 × 16 mm). Cell-free 3D matrix corresponding to a p150 plate was incubated with 1 × 108 exosomes per condition for 1 h at 37°C. After that, samples were ultracentrifuged at 10,000 g for 30 min to remove unbound cell-free 3D matrix components followed by an ultracentrifugation at 110,000 g for 70 min to recover exosomes. Obtained exosomes were loaded with equal amount of Laemmli buffer and analyzed by SDS-PAGE.
For assessing potential TnC internalization through endocytic trafficking, cells from the indicated genotypes were plated onto previously isolated decellularized CDMs (obtained as described above) and cultured for 24 h in the incubator at 37°C. Cells were then trypsinized and rinsed with PBS in order to remove any extracellular nonendocytosed ECM proteins. Finally, cells were resuspended into denaturing, reducing SDS-PAGE loading buffer to obtain “cell lysates” ready for Western blot analysis.
Lysates and exosomes derived from WT and Cav1−/− MEFs were digested using the filter-aided sample preparation protocol (Wiśniewski et al., 2009). Briefly, samples were dissolved in 50 mM Tris-HCl, pH 8.5, 4% SDS, and 50 mM DTT, boiled for 10 min, and centrifuged. Protein concentration in the supernatant was measured with the Direct Detect Spectrometer (Millipore). About 100 µg protein was diluted in 8 M urea in 0.1 M Tris-HCl, pH 8.5 (urea buffer; UA), and loaded onto 30-kD centrifugal filter devices (filter-aided sample preparation protein digestion kit). The denaturation buffer was removed by washing three times with UA. Proteins were then alkylated by incubation in 50 mM iodoacetamide in UA for 20 min in the dark, and excess alkylation reagents were eliminated by washing three times with UA and three additional times with 50 mM ammonium bicarbonate. Proteins were digested overnight at 37°C with modified trypsin (Promega) in 50 mM ammonium bicarbonate at a 50:1 protein:trypsin (wt/wt) ratio. The resulting peptides were eluted by centrifugation with 50 mM ammonium bicarbonate (twice) and 0.5 M sodium chloride. Trifluoroacetic acid was added to a final concentration of 1%, and the peptides were finally desalted onto C18 Oasis-HLB cartridges and dried down for further analysis.
Multiplexed isobaric labeling
For the quantitative analysis, tryptic peptides were dissolved in triethylammonium bicarbonate buffer, and the peptide concentration was determined by measuring amide bonds with the Direct Detect system. Equal amounts of each peptide sample were labeled using the 4-plex iTRAQ Reagents Multiplex Kit (Applied Biosystems). Briefly, each peptide solution was independently labeled at room temperature for 1 h with one iTRAQ reagent vial (mass tag 114, 115, 116, or 117) previously reconstituted with ethanol. After incubation at room temperature for 1 h, the reaction was stopped with diluted trifluoroacetic acid and peptides were combined. Samples were concentrated in a Speed Vac, desalted onto C18 Oasis-HLB cartridges, and dried down for mass spectrometry (MS) analysis.
Digested peptides were loaded into the liquid chromatography–tandem MS (LC-MS/MS) system for on-line desalting onto C18 cartridges and analyzed by LC-MS/MS using a C-18 reversed-phase nano-column (75 µm I.D. × 25 cm, 2 µm particle size, Acclaim PepMap RSLC, 100 C18; Thermo Fisher Scientific) in a continuous acetonitrile gradient consisting of 0%–30% B for 240 min and 50%–90% B for 3 min (A = 0.5% formic acid; B = 90% acetonitrile, 0.5% formic acid). A flow rate of 200 nl/min was used to elute peptides from the reverse phase nano-column to an emitter nanospray needle for real-time ionization and peptide fragmentation in an Orbitrap QExactive mass spectrometer (Thermo Fisher Scientific). During the chromatography run, we examined an enhanced Fourier transform-resolution spectrum (resolution = 70,000) followed by the higher-energy collisional dissociation MS/MS spectra from the 20 most intense parent ions. Dynamic exclusion was set at 40 s. For increased proteome coverage, labeled samples were also fractioned by cation exchange chromatography (Oasis HLB-MCX columns); fractions were desalted and analyzed using the same system and conditions described before.
Protein identification and quantification
All spectra were analyzed with Proteome Discoverer (version 126.96.36.199; Thermo Fisher Scientific) using SEQUEST-HT (Thermo Fisher Scientific). The Uniprot database, containing all mouse sequences (March 3, 2013), was searched with the following parameters: trypsin digestion with two maximum missed cleavage sites; precursor and fragment mass tolerances of 2 D and 0.03 D, respectively; methionine oxidation as a dynamic modification; and carbamidomethyl cysteine and N-terminal and Lys iTRAQ modifications as fixed modifications. Peptides were identified using the probability ratio method (Martínez-Bartolomé et al., 2008), and false discovery rate was calculated using inverted databases and the refined method (Navarro et al., 2014) with an additional filtering for a precursor mass tolerance of 15 ppm (Bonzon-Kulichenko et al., 2015). Identified peptides had a false discovery rate ≤1%.
Proteins were quantified from reporter ion intensities and quantitative data analyzed with QuiXoT, based on the weighted spectrum, peptide, and protein statistical model (García-Marqués et al., 2016). In this model, protein log2-ratios are expressed as standardized variables (i.e., in units of SD according to their estimated variances [Zq values]). Functional class enrichment was assessed using the Enrichr package (Chen et al., 2013), and network depiction of annotated relationships among hits was retrieved from the STRING v.11 resource).
MS raw files and msf search files have been deposited in the PeptideAtlas repository (accession no. PASS01614; http://www.peptideatlas.org/PASS/PASS01614).
LC-MS/MS analysis of exosome glycerophospholipids
Exosome samples from WT and KO mice corresponding to 2.5 µg protein were used. Before lipid extraction, the following internal standards were added: 200 pmol each of 1,2-dipentadecanoyl-sn-glycero-3-phosphocholine, 1,2-dilauroyl-sn-glycero-3-phosphoethanolamine, 1,2-dipalmitoyl-sn-glycero-3-phosphoinositol, 1,2-dimiristoyl-sn-glycero-3-phosphoserine, 1,2-dimiristoyl-sn-glycero-3-phosphate, and 1-heptadecanoyl-2-arachidonoyl-3-phosphoglycerol, according to the method of Bligh and Dyer (1959). After evaporation of organic solvent under vacuum, the lipids were redissolved in 50 µl solvent mixture (75% A, 25% B), and 40 µl was injected into an Agilent 1260 Infinity high-performance liquid chromatograph equipped with an Agilent G1311C quaternary pump and an Agilent G1329B Autosampler. The column was a FORTIS HILIC (150 × 3 mm, 3-µm particle size; Fortis Technologies) protected with a Supelguard LC-Si (20 mm × 2.1 mm) guard cartridge (Sigma). Glycerophospholipids and sphingomyelin were separated according to the procedure described by Axelsen and Murphy (2010), with minor modifications. Briefly, the mobile phase consisted of a gradient of solvent A (hexanes:isopropanol, 30:40, by vol) and solvent B (hexanes:isopropanol:20 mM AcONH4 in water, 30:40:7 by vol). The gradient started at 75% A from 0 to 5 min, then decreased from 75% A to 40% A at 15 min, from 40% A to 5% A at 20 min, holding at 5% until 40 min, and increasing to 75% at 41 min. The column was then reequilibrated by holding at 75% A for an additional 14 min before the next sample injection. The flow rate through the column was fixed at 0.4 ml/min, and this flow entered into the electrospray interface of an API2000 triple quadrupole mass spectrometer (Applied Biosystems). The source parameters were set as follows: ion spray voltage, −4,500 V; curtain gas, 20 psi; nebulizer gas, 35 psi; desolvation gas, 65 psi; and temperature, 400°C. Phospholipid species were analyzed in Q1 with negative ionization. Compound parameters were fixed for all analyzed species: declustering potential, −65 V; entrance potential, −10 V; focusing potential, −300 V; and collision cell exit potential, −10 V. Phospholipid species were detected as [M-H]− ions except for phosphatidylcholine and sphingomyelin species, which were detected as [M+OAc]− adducts. For quantification, chromatographic peaks of each species were integrated and compared with the peak area of the internal standard corresponding to each phospholipid class. Since bis(monoacylglycerol)phosphate and sphingomyelin internal standards were unavailable at the time of analysis, these species were quantified using 1-heptadecanoyl-2-arachidonoyl-3-phosphoglycerol and 1,2-dipentadecanoyl-sn-glycero-3-phosphocholine standards, respectively. MS source raw data have been deposited in the MetaboLights public repository under accession no. MTBLS1969.
Cholesterol quantification by LC-MS/MS
To quantify the amount of cholesterol in exosomes, 25 µl of 19.68 µg/ml cholesterol (d7) solution in methanol (SPLASH LipidoMIX Internal Standard; Avanti) was added to each sample. Cholesterol was extracted, and proteins were precipitated by adding 475 µl cold methanol–ethanol (1:1 vol/vol). Samples were vortex-mixed, sonicated during 3 min, and incubated on ice for 20 min. Finally, supernatants were collected by centrifugation at 18,400 g for 20 min at 8°C and dried in SpeedVac (Savant SPD131DDA concentrator; Thermo Fisher Scientific) for 2 h at room temperature.
Metabolomics analysis was performed using an Ultimate 3000 HPLC system coupled to an Orbitrap ELITE Hybrid Ion Trap-Orbitrap Mass Spectrometer (Thermo Fisher Scientific). For chromatographic separation, an Agilent RP, Hi Recovery column (100 × 0.5 mm, 5 µ) was used. Cholesterol was identified by LC-MS/MS mode and quantified according to the response factor of the respective internal standard.
Intracellular compartments were separated by sucrose density gradient centrifugation. Cells were washed three times in 10 mM Tris-HCl, pH 7.5, and 0.5 mM MgCl2 and then scraped and resuspended in 50 mM Tris-HCl, pH 7.5, and 0.25 M sucrose. Cells were homogenized with 30 strokes using a Dounce homogenizer. The homogenates were then centrifuged at 3,000 g for 15 min at 4°C. The supernatant was placed over a continuous sucrose gradient (2 M at the bottom; 0.25 M at the top), and samples were separated at 100,000 g for 18 h in an SW40 rotor (Beckman Coulter) at 4°C. After centrifugation, 1-ml fractions were collected from the tube, yielding a total of 12 fractions. Cold acetone (1 ml) was added to each tube, and the homogeneous mixture was allowed to precipitate at −20°C for 2 h. The samples were centrifuged at 16,000 g at 4°C, and the protein pellets were allowed to dry for 2 h. The protein precipitates were then analyzed by SDS-PAGE for proteins, including the ER marker calreticulin and the MVB marker CD63.
Preparation of DRM fractions
DRMs from exosomes derived from Cav1WT and Cav1KO MEFs were purified as described in Navarro-Lérida et al. (2002), with the Triton X-100 concentration reduced from 1% to 0.5%.
Transwell migration assay
The transwell migration assay was performed with Boyden chambers containing an 8-µm pore-size polycarbonate filter (Corning Costar). Serum-free medium (300 µl) with or without exosomes was placed in the wells of a 24-well plate. The transwell inserts were placed in the wells, and 5 × 104 MDA-MB468 cells in 200 µl serum-free medium were added to the upper chamber. After incubation for 24 h, cells on the lower membrane surface were fixed and stained with crystal violet and counted under a microscope.
All animal procedures conformed to European Union Directive 86/609/EEC and Recommendation 2007/526/EC regarding the protection of animals used for experimental and other scientific purposes, enforced in Spanish law under Real Decreto 1201/2005. Cav1WT and Cav1−/− exosomes were labeled with PKH67 and intravenously injected every day for 1 wk into the tail vein of TnC−/− mice (Talts et al., 1999) bred into the FVB background. After 4 d, lungs and livers were collected and embedded sections were analyzed by immunohistochemistry with anti-TnC antibody (Sigma).
Cells grown on glass coverslips were washed twice with PBS and fixed for 15 min at room temperature in 4% PFA in PBS. Fixed cells were washed extensively with PBS and permeabilized for 5 min in PBS containing 0.1% TritonX-100 and 2% BSA to reduce nonspecific binding. Cells were incubated for 1 h at room temperature in PBS containing 0.2% BSA with primary antibodies (typically diluted 1:200). Subsequent washes and incubations with Alexa Fluor 647 or 546 phalloidin or fluorescent secondary antibodies were performed in PBS for 1 h at room temperature. Coverslips were mounted in PermaFluor aqueous mounting medium and examined with a Leica SPE, SP5, or SP8 confocal microscope, typically through 40× (ACS APO 40.0× 1.15 oil) or 63× (ACS APO 63.0× 1.30 oil) objectives. A 100× objective (HC PL Apo CS2 100×/NA 1.4 oil) was used for high-resolution imaging in an SP8 confocal microscope.
Filipin staining and quantification
Cells were grown on glass coverslips. After fixing for 10 min at 37°C with 4% PFA in PBS, cells were permeabilized with 0.2% BSA and 0.1% saponin in PBS (blocking solution) for 30 min. Samples were then incubated with primary antibody (LBPA) and fluorescently labeled–conjugated secondary antibodies for 1 h each, interspersed with three washes in blocking solution. After an additional wash in PBS, cells were incubated for 1 h with filipin (50 µg/ml in PBS). After three washes in PBS, slides were mounted in Fluoromount. Fiji software was used to quantify filipin fluorescence intensity (http://fiji.sc/). Cells were segmented based on LBPA staining. Filipin staining intensity in LBPA-positive vesicles was measured after intensity thresholding.
Quantitative real-time PCR
Frozen tissues or cell lines were analyzed for specific gene expression using Sybr Green PCR Reagents (Applied Biosystems) and specific primers: mouse Tsg101 sense 5′-TCTAACCGTCCGTCAAACTGT-3′, antisense 5′-TTGTACCAGTGAGGTTCACCA-3′; mouse nuclear sphingomyelinase (nSMase) 2 sense 5′-ACACGACCCCCTTTCCTAATA-3′, antisense 5′-GGCGCTTCTCATAGGTGGTG-3′; and mouse nSMase 1 sense 5′-TGGGACATCCCCTACCTGAG-3′, antisense 5′-TAGGTGAGCGATAGCCTTTGC-3′. Total RNA was extracted from tissues or cells using the RNeasy kit (Qiagen) and reverse-transcribed. Quantitative real-time PCR was performed with a 7500Fast Real-Time PCR System. Relative expression was normalized to GAPDH and hypoxanthine guanine phosphoribosyl transferase.
Harvested tissues were fixed in 4% PFA, processed through graded sucrose, embedded in O.C.T. medium (Tissue-Tek), and stored at −80°C. Frozen sections (20 µm) were dried for 10 min at room temperature and blocked for 1 h in PBS containing 4% chicken serum and 0.3% TritonX-100. Sections were stained with anti-TnC antibody diluted 1:200 in blocking solution overnight at 4°C. Samples were washed extensively in PBS, 0.15% TritonX-100 for at least for 3 h. Samples were then incubated overnight with anti-rabbit Alexa Fluor 647 secondary antibody (1:200), Alexa Fluor 546 phalloidin (1:500), and Hoechst 33342 in blocking solution. After extensive washing, sections were mounted in PermaFluor mounting medium. Confocal Z-stack images were captured, and TnC deposits were quantified.
Extracellular TnC and FN fiber quantification
The procedure for image processing and quantification of “fiberness” was developed as a macro in Fiji (ImageJ 1.50e x64). Fiberness measures the amount of fiber-like structures in an image, providing a readout that takes into account both the density of fibers and their length independently of orientation. Noise reduction first stabilizes the image background. The structural information from the eigen values of the Hessian matrix is then used to apply a Frangi vesselness filter76 to enhance very thin, almost unidimensional tubular structures (filter scale: σ = 0.179 µm, close to pixel size). The output is a fiber-enhanced image in which each pixel contains a fiberness score. M0 readout was computed as the mean fiberness score in the whole image. The processing pipeline and extracted M0 measures are illustrated in Fig. S3 B.
LDL incorporation in vitro
Blood for LDL isolation was obtained from healthy volunteers after informed consent in accordance with local ethics committee guidelines (Comité de Ética de la Investigación del Instituto de Salud Carlos III: CEI PI 12_2016-v2).
Fluorescently labeled human LDL was prepared as described (Steffensen et al., 2015). Briefly, human blood from healthy donors was collected in K3EDTA-containing tubes. After centrifugation, plasma was collected, mixed with KBr, and layered on a KBr density gradient column. The column was centrifuged at 256,000 g for 18 h in an Optima L-100 ultracentrifuge (Beckman Coulter). The 1.063-g/ml density layer containing LDL was collected, purified on a PD10 column (17085101; GE Healthcare), and conjugated to the fluorochrome Alexa Fluor 647. Alexa Fluor 647–LDL was purified on a PD10 column, and purity was tested at the Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC) Proteomics Unit by LC-MS.
Cav1WT MEFs were cultured in DMEM supplemented with 2% FBS and loaded with Alexa Fluor 647–LDL to a final concentration of 200 µg/ml for 24 h (for TnC analysis at MVBs) or for 48 h (for TnC deposition studies).
Error bars depict SEM. Statistical significance was determined with GraphPad Prism by unpaired Student’s t test or Mann-Whitney test, as indicated; *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Online supplemental material
Fig. S1 shows that internalization of Cav1 favors its entry into exosomes. Fig. S2 shows that Cav1-positive exosomes display a specific lipid/protein composition. Fig. S3 shows that exosome inhibition favors intracellular TnC accumulation. Fig. S4 shows that exosomes derived from CAFs modulate ECM deposition. Fig. S5 shows that disruption of ER-MVB contact sites modulates TnC matrix deposition. Data S1 presents high-throughput comparative proteomics of cell lysates and exosomes derived from either WT or Cav1KO fibroblasts. Data S2 presents high-throughput comparative proteomics of exosomes derived from MDA-MB231 or MDA-MB468 cells.
We thank the CNIC Microscopy and Histology Units for technical assistance. We thank Daniel Jiménez-Carretero (CNIC Cellomics Unit) for developing the software for TnC fiber quantification, Dr. Kirchner of Centre for Medical Biotechnology (University of Duisburg-Essen, Duisburg, Germany) for plasmids encoding ubiquitinated Cav1 mutants, Fátima Sánchez-Cabo (CNIC Bioinformatics Unit) for statistical analysis of exosome populations, Dr. Germán Andrés (Centro de Biologia Molecular Severo Ochoa) for help with electron microscopy, Dr. Fidel Lolo (CNIC, Madrid, Spain) for PTRF-reconstituted PTRFKO cells, Mauro Catalá for technical assistance, and Giulio Fulgoni (CNIC, Madrid, Spain) for fluorescently labeled LDLs. Miguel Sánchez-Álvarez provided help with manuscript editing.
This study was supported by the Ministerio de Ciencia, Innovación y Universidades (CSD2009-0016, SAF2014-51876-R, SAF2017-83130-R, BFU2016-81912-REDC, and IGP-SO-MINSEV1512-07-2016); the Fundació La Marató de TV3 (385/C/2019); the Worldwide Cancer Research Foundation (AICR 15-0404); and Fondo Europeo de Desarrollo Regional “Una manera de hacer Europa” (to M.Á. del Pozo). M.Á. del Pozo’s group received funding from the European Union Horizon 2020 research and innovation program under Marie Sklodowska-Curie grant agreement no. 641639. M.Á. del Pozo is a member of the Tec4Bio consortium (ref. S2018/NMT4443; Actividades de I+D entre Grupos de Investigación en Tecnologías, Comunidad Autónoma de Madrid/FEDER, Spain). J. Balsinde was supported by Ministerio de Ciencia, Innovación y Universidades grants SAF2013-48201-R and SAF2016-80883-R, and G. Orend was supported by Institut National de la Santé et de la Recherche Médicale, the University of Strasbourg, the Ligue Contre le Cancer, and the Institut National du Cancer (ref. TENPLAMET). L. Albacete-Albacete was supported by a Ministerio de Ciencia, Innovación y Universidades predoctoral fellowship associated with the Severo Ochoa Excellence program (ref. SVP-2013-06789). I. Navarro-Lérida was supported by a postdoctoral fellowship from the Asociación Española Contra el Cáncer (ref. INVES191NAVA). The Centro Nacional de Investigaciones Cardiovasculares Carlos III is supported by the Ministerio de Ciencia e Innovación and the Pro CNIC Foundation and is a Severo Ochoa Center of Excellence (SEV-2015-0505).
The authors declare no competing financial interests.
Author contributions: I. Navarro-Lérida and M.Á. del Pozo Barriuso conceived and designed the project and wrote the manuscript. L. Albacete-Albacete and I. Navarro-Lérida performed most of the experiments and analyzed and interpreted the data. L. Albacete-Albacete prepared the manuscript figures. J.A. López and J. Vázquez performed and analyzed proteomic studies. A.M. Astudillo, J. Balsinde, and A. Ferrarini performed and analyzed lipidomic experiments. I. Martín-Padura helped with in vivo experiments. G. Orend and M. Van-Der-Heyden provided reagents, TnCKO animals essential for the study, and scientific feedback.